Redis Mutex

Distributed mutex in Ruby using Redis. Supports both blocking and non-blocking semantics.

The idea was taken from the official SETNX doc.

In the following example, only one thread / process / server can enter the locked block at one time.

RedisMutex.with_lock(:your_lock_name) do
  # do something exclusively
end

or

mutex = RedisMutex.new(:your_lock_name)
if mutex.lock
  # do something exclusively
  mutex.unlock
else
  puts "failed to acquire lock!"
end

By default, while one is holding a lock, others wait 1 second in total, polling every 100ms to see if the lock was released. When 1 second has passed, the lock method returns false and others give up. Note that if your job runs longer than 10 seconds, the lock will be automatically removed to avoid a deadlock situation in case your job is dead before releasing the lock. Also note that you can configure any of these timing values, as explained later.

Or if you want to immediately receive false on an unsuccessful locking attempt, you can change the mutex mode to non-blocking.

redis-mutex 4.0 has brought a few backward incompatible changes to follow the major upgrade of the underlying redis-classy gem.

• The base class Redis::Mutex is now RedisMutex.
• Redis::Classy.db = Redis.new is now RedisClassy.redis = Redis.new.

• Ruby 2.0 or later is required.
• auto_mutex now takes :on for additional key scoping.

• Exception-based control flow: Added lock! and unlock!, which raises an exception when fails to acquire a lock. Raises RedisMutex::LockError and RedisMutex::UnlockError respectively.
• INCOMPATIBLE CHANGE: #lock no longer accepts a block. Use #with_lock instead, which uses lock! internally and returns the value of block.
• unlock returns boolean values for success / failure, for consistency with lock.

gem install redis-mutex

In Gemfile:

gem 'redis-mutex'

Register the Redis server: (e.g. in config/initializers/redis_mutex.rb for Rails)

RedisClassy.redis = Redis.new

Note that Redis Mutex uses the redis-classy gem internally to organize keys in an isolated namespace.

There are a number of methods:

mutex = RedisMutex.new(key, options)    # Configure a mutex lock
mutex.lock                                # Try to acquire the lock, returns false when failed
mutex.lock!                               # Try to acquire the lock, raises exception when failed
mutex.unlock                              # Try to release the lock, returns false when failed
mutex.unlock!                             # Try to release the lock, raises exception when failed
mutex.locked?                             # Find out if resource already locked
mutex.with_lock                           # Try to acquire the lock, execute the block, then return the value of the block.
                                          # Raises exception when failed to acquire the lock.

RedisMutex.sweep                        # Remove all expired locks
RedisMutex.with_lock(key, options)      # Shortcut to new + with_lock

The key argument can be symbol, string, or any Ruby objects that respond to id method, where the key is automatically set as TheClass:id. For any given key, RedisMutex: prefix will be automatically prepended. For instance, if you pass a Room object with id of 123, the actual key in Redis will be RedisMutex:Room:123. The automatic prefixing and instance binding is the feature of RedisClassy - for more internal details, refer to Redis Classy.

The initialize method takes several options.

:block  => 1    # Specify in seconds how long you want to wait for the lock to be released.
                # Specify 0 if you need non-blocking sematics and return false immediately. (default: 1)
:sleep  => 0.1  # Specify in seconds how long the polling interval should be when :block is given.
                # It is NOT recommended to go below 0.01. (default: 0.1)
:expire => 10   # Specify in seconds when the lock should be considered stale when something went wrong
                # with the one who held the lock and failed to unlock. (default: 10)

The lock method returns true when the lock has been successfully acquired, or returns false when the attempts failed after the seconds specified with :block. When 0 is given to :block, it is set to non-blocking mode and immediately returns false.

In the following Rails example, only one request can enter to a given room.

class RoomController < ApplicationController
  before_filter { @room = Room.find(params[:id]) }
  
  def enter
    RedisMutex.with_lock(@room) do    # key => "Room:123"
      # do something exclusively
    end
    render text: 'success!'
  rescue RedisMutex::LockError
    render text: 'failed to acquire lock!'
  end
end

Note that you need to explicitly call the unlock method when you don't use with_lock and its block syntax. Also it is recommended to put the unlock method in the ensure clause.

def enter
  mutex = RedisMutex.new('non-blocking', block: 0, expire: 10.minutes)
  if mutex.lock
    begin
      # do something exclusively
    ensure
      mutex.unlock
    end
    render text: 'success!'
  else
    render text: 'failed to acquire lock!'
  end
end

If you want to wrap an entire method into a critical section, you can use the macro-style definition. The locking scope will be TheClass#method and only one method can run at any given time.

If you give a proc object to the after_failure option, it will get called after locking attempt failed.

class JobController < ApplicationController
  include RedisMutex::Macro
  auto_mutex :run, block: 0, after_failure: lambda { render text: 'failed to acquire lock!' }

  def run
    # do something exclusively
    render text: 'success!'
  end
end

Also you can specify method arguments with the on option. The following creates a mutex key named ItunesVerifier#perform:123456, so that the same method can run in parallel as long as the transaction_id is different.

class ItunesVerifier
  include Sidekiq::Worker
  include RedisMutex::Macro
  auto_mutex :perform, on: [:transaction_id]

  def perform(transaction_id)
    ...
  end
end